From glycerol to allyl alcohol: iron oxide catalyzed dehydration

and consecutive hydrogen transferw

Yong Liu, Harun Tuysuz, Chun-Jiang Jia, Manfred Schwickardi,

Roberto Rinaldi, An-Hui Lu, Wolfgang Schmidt and Ferdi Schuth*

Received (in Cambridge, UK) 16th October 2009, Accepted 21st December 2009

First published as an Advance Article on the web 18th January 2010

DOI: 10.1039/b921648k

Using iron oxide as catalyst, glycerol can be converted to allyl

alcohol through a dehydration and consecutive hydrogen transfer.

Due to decreasing petroleum reserves and increasing energy

demand, biomass, as a sustainable and renewable alternative

source, has attracted great interest.1 Biodiesel is one of the few

examples to have been produced on a large scale and production

will continue to increase in the coming years.2 Thus, as the

byproduct of biodiesel production, glycerol is considered as a

potential biorefinery feedstock. To date, various processes

have been developed to obtain valuable chemicals from glycerol,

such as reforming,3 oxidation,4 hydrogenolysis,5 etherification,6

and dehydration.7 For the selective dehydration of glycerol to

acrolein, catalysts with strong acidity are commonly used and

acrolein yields of up to 75% can be obtained. However, the

catalysts often suffer from fast deactivation.7

The selective hydrogenation of acrolein to allyl alcohol is of

great interest from both the industrial and academic point of

view.8 However, since the hydrogenation of the CQC double

bond is strongly favored over hydrogenation of the CQO

double bond, only in a few cases could allyl alcohol be highly

selectively produced by the hydrogenation of acrolein. One

possibility to selectively hydrogenate the CQO double bond is

the use of supported silver9 or gold10 catalysts; another

possibility is to go through a gas-phase hydrogen transfer

reaction which is similar to the Meerwein–Ponndorf–Verley

reduction.11 Recently, we reported the pseudomorphic trans-

formation of mesoporous Co3O4 and ferrihydrite to CoO and

Fe3O4.12 Now we have studied this system from the catalytic

point of view, and discovered that both materials can catalyze

not only the dehydration of glycerol to acrolein, but also

consecutively the transfer hydrogenation to allyl alcohol,

albeit presently still in only moderate yields. While this

observation alone is interesting, the mesostructured oxides

are obtained only after a complex synthesis sequence, and

thus alternative, cheaper catalysts would be needed. Iron oxide

would be preferred over cobalt oxide, since it is less toxic and

cheaper. In addition, a simpler synthetic access to an active

catalyst is required. Herein we report a simple synthesis of

iron oxide with high surface area, over which almost full

conversion of glycerol and 20–25% yield of allyl alcohol could

be achieved. The selectivity in the second step, the transfer

hydrogenation to allyl alcohol, is close to 100%, and only

slight deactivation was observed after 72 h.

Phosphorus containing Fe2O3 nanoparticles were syn-

thesized through a very simple procedure by adding a small

amount of Al(H2PO4)3 to diluted Fe(NO3)3 aqueous solution

and then heating the mixture up to 380 1C for 1 h (experi-

mental details are available in the Supporting Informationw)until dryness. Afterwards Fe2O3 nanoparticles with a nearly

amorphous structure and a surface area of 164 m2 g�1 could be

collected (ESIw, Fig. S2 and S3). The addition of Al(H2PO4)3 is

necessary to obtain high surface areas; without phosphate addi-

tion Fe2O3 materials with much lower and poorly reproducible

BET surface areas (varying between 47 and 85 m2 g�1 in three

different batches) were obtained. The catalytic performance of

phosphate containing Fe2O3 with high surface area in glycerol

conversion was tested, and the result is shown in Fig. 1a. It can

be seen that at a space velocity (GHSV) of 96.4 h�1, full

conversion of glycerol could be achieved in the first 6 h, and

after 24 h the conversion still remains at around 95%, which is

quite remarkable compared with former reports. However,

even more interesting is the fact that allyl alcohol was formed

as the main product, besides acrolein and hydroxyacetone,

although no molecular hydrogen was present in the system.

The unusually high and unexpected yield of allyl alcohol is

very interesting, because allyl alcohol is the product of the

selective hydrogenation of the CQO double bond of acrolein,

while propanal or even propanol would be expected to form as

the product of CQC double bond hydrogenation. Acrolein,

hydroxyacetone and allyl alcohol were quantitatively analyzed,

and the results are shown in Fig. 1b. It can be seen that the

distribution of the products is quite stable over reaction time

and 6–9% of acrolein, 18–20% of hydroxyacetone, and

20–25% of allyl alcohol could be obtained, which add up to

50–60% of the total carbon balance. Trace amounts of

Fig. 1 (a) Glycerol conversion and (b) product distribution at 320 1C

with a space velocity of 96.4 h�1.

Max-Planck-Institut fur Kohlenforschung, Kaiser-Wilhelm-Platz 1,45470 Mulheim an der Ruhr, Germany.E-mail: schueth@mpi-muelheim.mpg.de; Fax: +49 208-306-2995w Electronic supplementary information (ESI) available: Detailedexperimental procedures, measurement conditions and additionalcatalytic data. See DOI: 10.1039/b921648k

1238 | Chem. Commun., 2010, 46, 1238–1240 This journal is �c The Royal Society of Chemistry 2010

COMMUNICATION www.rsc.org/chemcomm | ChemComm

Dow

nloa

ded

by M

ount

Alli

son

Uni

vers

ity o

n 09

/05/

2013

01:

12:0

2.

Publ

ishe

d on

18

Janu

ary

2010

on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/B

9216

48K

View Article Online / Journal Homepage / Table of Contents for this issue

propanal, acetone and acetaldehyde were also detected by

GC-MS. However, since the signal intensities are very low,

these species were not calibrated (detailed discussion about the

total carbon balance is available in the ESIw). Since only a verysmall amount of propanal was formed and propanol was not

detected at all, it can be concluded that iron oxide can not only

catalyze the dehydration of glycerol to acrolein, but also

consecutively the selective hydrogenation of acrolein to

allyl alcohol.

Studies on the possible mechanism for the formation of allyl

alcohol were then carried out. Bergman et al. recently reported

the preparation of allyl alcohol via a formic acid-mediated

deoxygenation of glycerol.13 However, stoichiometric

amounts of formic acid are needed in that case, and this

mechanism is highly unlikely for the current system, since

formic acid was never detected in our experiments. The

probable reaction sequence is shown in Scheme 1. Hydroxy-

acetone and acrolein are both products of the dehydration of

glycerol, and allyl alcohol could be obtained either through the

direct hydrogenation of acrolein with molecular hydrogen, or

through a hydrogen transfer reaction with an alcohol (glycerol

or some intermediates in this case) as hydrogen donor. The

possibility of direct hydrogenation was excluded by two facts.

Firstly, no molecular hydrogen was supplied to the system;

secondly, pure acrolein instead of glycerol was pumped

through the reactor under the same conditions with and without

hydrogen, and no allyl alcohol was detected in either case.

Catalysts such as MgO, ZrO2 and MgAl2O4 are known to

be active for hydrogen transfer reactions,11 while iron oxide

has never been reported for this type of reaction. Therefore,

several reactions between different alcohols and a,b-unsaturatedaldehydes were carried out and the results are shown in

Table 1. It can be seen that hydrogen transfer reactions can

indeed take place in all cases, although the yields of the allylic

alcohols are considerably lower than for the iron oxide–

glycerol system. It seems that the hydrogen transfer does not

only include the reaction between glycerol and acrolein.

Instead, probably intermediates with hydroxy groups forming

during the reaction play an important role, and these inter-

mediates together with glycerol then react as hydrogen donor

with acrolein. Further research on a more detailed mechanism

is still ongoing in our laboratory, although this is rather

difficult due to the complexity of the system. However, it is

clear that iron oxide plays a rather unique role, even when

compared with the already known hydrogen transfer catalysts.

For comparison, MgO and MgAl2O4 with high surface areas

were synthesized and used for both glycerol conversion and

the reaction between glycerol and acrolein. In all cases,

the yield of allyl alcohol was below 1% (the preparation

procedures and properties of the MgO andMgAl2O4 materials

are supplied in the ESIw).To better understand the system, studies on the structure

and components of the catalyst before and after the reaction

were carried out. The catalyst used in the current system is

quite complicated. On the one hand, according to our previous

report, the catalyst should be reduced from Fe2O3 to Fe3O4

during the reaction.12 On the other hand, to obtain high

surface area, Al(H2PO4)3 was added which probably leads to

the formation of Al2O3 and phosphates. As far as the

structural change of the catalyst during the reaction is con-

cerned, XRD analysis showed that the reduction of Fe2O3 to

Fe3O4 also took place in the current system during reaction

(ESIw, Fig. S4) and further studies confirmed that the catalyst

reduction procedure was already finished after 6 h. The

broadening of the XRD reflections indicates that the size of

the catalyst is still very small. However, nitrogen sorption

measurements showed that after 24 h, the BET surface area of

the catalyst had dropped significantly from 164 to 16 m2 g�1.

This could be caused by the carbon depositions on the surface

of the catalyst. High glycerol conversion was observed before

and after the reduction of Fe2O3 to Fe3O4, however one

cannot compare the activities of Fe2O3 and Fe3O4 based on

the data given in Fig. 1a, since at 100% conversion no

difference could be seen any more. The experiment was then

carried out at the same temperature (320 1C) but at increased

GHSV of 413 h�1 instead of 96.4 h�1, and the results are

shown in ESIw, Fig. S5. It can be seen that the glycerol

conversion dropped to 30–40%. However, no deactivation

could be observed over 24 h, which indicates that the catalyst

Scheme 1 Possible reaction pathways from glycerol to allyl alcohol.

Table 1 Yields of hydrogen transfer reactions between differenta,b-unsaturated aldehydes and alcohols catalyzed by iron oxide

Entry Substrate H-donor Product Yield [%]a

1b1.57

2b1.36

3b 4.50

4b2.18

5b6.50

6c16.90

7d1.72

8d3.87

a All yields are based on GC results with quantification. b Acrolein

was premixed with the corresponding alcohol at a molar ratio of 1 : 1

and then pumped through the reactor at the same temperature, catalyst

amount and flow rate. c Glycerol : acrolein molar ratio = 2.4 : 1,

allyl alcohol yield was calculated based on the amount of glycerol.d Crotonaldehyde and cinnamaldehyde were mixed with 50 wt%

glycerol by a supersonic bath since they are not miscible with glycerol

(detailed procedures are given in ESIw).

This journal is �c The Royal Society of Chemistry 2010 Chem. Commun., 2010, 46, 1238–1240 | 1239

Dow

nloa

ded

by M

ount

Alli

son

Uni

vers

ity o

n 09

/05/

2013

01:

12:0

2.

Publ

ishe

d on

18

Janu

ary

2010

on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/B

9216

48K

View Article Online

has very good stability and both Fe2O3 and Fe3O4 can

effectively catalyze this reaction. To further confirm this, the

catalyst after the reaction (already reduced to Fe3O4) was

reused at 320 1C with a GHSV of 96.4 h�1. Glycerol con-

versions of 80–90% were reached over 24 h and no deactivation

was observed (ESIw, Fig. S6). In the solid acid catalyst

catalyzed glycerol dehydration to acrolein, the deactivation

of the catalyst is usually attributed to carbon deposition on the

surface of the catalyst.7 For the catalyst system described here,

thermogravimetric (TG) analysis of the catalyst after 24 h

reaction time (320 1C, GHSV 96.4 h�1) revealed a weight loss

of 22.8% (ESIw, Fig. S7a). In interpreting this result, one also

needs to take into account that the catalyst which was

collected after reaction had already been reduced to Fe3O4.

During the TG measurement, it would be re-oxidized to

g-Fe2O3 (as also confirmed by DSC results). Taking the

3.7% weight gain resulting from the oxidation into account,

a total weight loss of 26.5% can be calculated. This might

partly explain why the nitrogen sorption measurements

showed that after 24 h, the BET of the catalyst dropped

significantly from 164 to 16 m2 g�1. The broadening of

XRD reflections of the material (ESIw, Fig. S4b) indicates

that the particle size is still small, and the reduction in surface

area is thus probably caused by pore blocking. When the

reaction time was further prolonged to 72 h (320 1C, GHSV

96.4 h�1), the weight loss increased to 31.6% (ESIw, Fig. S7b),and accordingly a reduced glycerol conversion of 90.4% and

allyl alcohol yield of 16.3% were observed.

As far as the presence of Al2O3 and phosphates is

concerned, one would expect that their acidity would have

an effect on the formation of acrolein and consequently on the

whole reaction, since the dehydration of glycerol to acrolein is

normally catalyzed by catalysts with strong Brønsted acidity.

To confirm this, another method was used to synthesize Fe2O3

nanoparticles with reproducible BET surface area (85 m2 g�1)

without the addition of Al(H2PO4)3 (detailed synthesis pro-

cedure is available in the ESIw). Glycerol conversion and

product distribution are shown in ESIw Fig. S8 and Fig. S9,

respectively. Glycerol conversions of close to 100% and allyl

alcohol yields of around 25% could be observed in the first

6 h, indicating that iron oxide is the active species and the

addition of Al(H2PO4)3 in principle does not play a crucial

role. The Brønsted acidity needed for the glycerol dehydration

could then come from both the hydroxy groups on the surface

of iron oxide and the autoprotolysis of water on the surface

defects of iron oxide at a temperature as high as 320 1C. Since

the Brønsted acid sites are not as strong as in typical solid acid

catalysts such as zeolites and sulfated zirconia, the conversion

of glycerol is also lower. This lower activity is actually

advantageous for this system since unreacted glycerol may

be needed as the hydrogen donor in the following hydrogen

transfer reaction. However, it is noteworthy that without the

addition of Al(H2PO4)3, a more significant deactivation of the

catalyst was observed (glycerol conversion dropped to 80%

after 24 h and allyl alcohol yield decreased to 20%). However,

the reason for the better stability of the phosphorus containing

sample is still not clear.

In summary, a process for the conversion of glycerol to allyl

alcohol over an iron oxide catalyst was discovered. A dehydration

and consecutive hydrogen transfer mechanism is proposed,

although the details of how the hydrogen transfer takes place

are still under investigation. The iron oxide catalyst shows high

activity and good stability, and due to its low price, easily tunable

structures and morphologies, further improvements can be

expected which may even open the route to practical applications.

Notes and references

1 G. W. Huber, S. Iborra and A. Corma, Chem. Rev., 2006, 106,4044; G. W. Huber, J. N. Chheda, C. J. Barrett and J. A. Dumesic,Science, 2005, 308, 1446; Y. Roman-Leshkov, C. J. Barrett,Z. Y. Liu and J. A. Dumesic, Nature, 2007, 447, 982; R. Rinaldi,R. Palkovits and F. Schuth, Angew. Chem., 2008, 120, 8167;R. Rinaldi, R. Palkovits and F. Schuth, Angew. Chem., Int. Ed.,2008, 47, 8047; R. Rinaldi and F. Schuth, Energy Environ. Sci.,2009, 2, 610.

2 R. G. Bray, Biodiesel Production, SRI Consulting, USA, 2004.3 R. R. Soares, D. A. Simonetti and J. A. Dumesic, Angew. Chem.,2006, 118, 4086; R. R. Soares, D. A. Simonetti and J. A. Dumesic,Angew. Chem., Int. Ed., 2006, 45, 3982; D. A. Simonetti, J. Rass-Hansen, E. L. Kunkes, R. R. Soares and J. A. Dumesic, GreenChem., 2007, 9, 1073.

4 S. Carrettin, P. McMorn, P. Johnston, K. Griffin andG. J. Hutchings, Chem. Commun., 2002, 696; F. Porta and L. Prati,J. Catal., 2004, 224, 397; R. Ciriminna, G. Palmisano, C. D. Pina,M. Rossi and M. Pagliaro, Tetrahedron Lett., 2006, 47, 6993.

5 M. A. Dasari, P. P. Kiatsimkul, W. R. Sutterlin and G. J. Suppes,Appl. Catal., A, 2005, 281, 225.

6 A. M. Ruppert, J. D. Meeldijk, B. W. M. Kuipers, B. H. Erne andB. M. Weckhuysen, Chem.–Eur. J., 2008, 14, 2016.

7 L. Ott, M. Bicker and H. Vogel, Green Chem., 2006, 8, 214;A. Corma, G. W. Huber, L. Sauvanauda and P. O’Connor,J. Catal., 2008, 257, 163; S. H. Chai, H. P. Wang, Y. Liang andB. Q. Xu, Green Chem., 2007, 9, 1130; H. Atia, U. Armbruster andA. Martin, J. Catal., 2008, 258, 71; C.-J. Jia, Y. Liu, W. Schmidt,A.-H. Lu and F. Schuth, J. Catal., 2010, 269, 71.

8 P. Claus, Top. Catal., 1998, 5, 51; P. Gallezot and D. Richard,Catal. Rev. Sci. Eng., 1998, 40, 81.

9 C. E. Volckmar, M. Bron, U. Bentrup, A. Martin and P. Claus,J. Catal., 2009, 261, 1; M. Bron, D. Teschner, A. Knop-Gericke,B. Steinhauer, A. Scheybal, M. Havecker, D. Wang, R. Fodisch,D. Honicke, A. Wootsch, R. Schlogl and P. Claus, J. Catal., 2005,234, 37; M. Bron, D. Teschner, A. Knop-Gericke, F. C. Jentoft,J. Krohnert, J. Hohmeyer, C. Volckmar, B. Steinhauer, R. Schlogland P. Claus, Phys. Chem. Chem. Phys., 2007, 9, 3559.

10 J. E. Bailie and G. J. Hutchings, Chem. Commun., 1999, 2151;J. E. Bailie, H. A. Abdullah, J. A. Anderson, C. H. Rochester,N. V. Richardson, N. Hodge, J.-G. Zhang, A. Burrows, C. J. Kielyand G. J. Hutchings, Phys. Chem. Chem. Phys., 2001, 3, 4113;P. Claus, A. Brueckner, C. Mohr and H. Hofmeister, J. Am. Chem.Soc., 2000, 122, 11430; C. Mohr, H. Hofmeister, J. Radnik andP. Claus, J. Am. Chem. Soc., 2003, 125, 1905; C. Mohr,H. Hofmeister and P. Claus, J. Catal., 2003, 213, 86.

11 S. A. Ballard, H. D. Finch and D. E. Winkler, Adv. Catal., 1957, 9,754; M. Ueshima and Y. Shimasaki, Catal. Lett., 1992, 15, 405;F. J. Urbano, M. A. Aramendıa, A. Marinas and J. M. Marinas,J. Catal., 2009, 268, 79; F. Braun and J. I. D. Cosimo, Catal.Today, 2006, 116, 206.

12 H. Tuysuz, Y. Liu, C. Weidenthaler and F. Schuth, J. Am. Chem.Soc., 2008, 130, 14108.

13 E. Arceo, P. Marsden, R. G. Bergman and J. A. Ellman, Chem.Commun., 2009, 3357.

1240 | Chem. Commun., 2010, 46, 1238–1240 This journal is �c The Royal Society of Chemistry 2010

Dow

nloa

ded

by M

ount

Alli

son

Uni

vers

ity o

n 09

/05/

2013

01:

12:0

2.

Publ

ishe

d on

18

Janu

ary

2010

on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/B

9216

48K

View Article Online